Palettes and bad rendering

Lee Daniel Crocker (lcrocker@calweb.com)
Fri, 3 May 1996 15:15:57 -0700 (PDT)


Message-Id: <199605032215.PAA25716@web1.calweb.com>
Subject: Palettes and bad rendering
To: www-html@w3.org
Date: Fri, 3 May 1996 15:15:57 -0700 (PDT)
From: "Lee Daniel Crocker" <lcrocker@calweb.com>

The issue is that the original poster ran into what he preceived
as a problem, and suggested a possible (though granted, very bad)
solution.  Arguing about how bad it is or whether we should shuffle
it over to HTTP or CSS isn't focusing on the problem in the first
place.  Let's step back a minute:

The problem, as I see it: When one designs a presentation with
many spot colors, and it is displayed by most browsers running on
palette-based displays, it looks like crap, even if the display
in question is as capable as that on which it was designed.

As I see it, then, the issue is not one of lack of information--
sending more information, therefore, won't help.  The issue is
that browsers are getting the information and doing a lousy job
displaying it.  Each image already contains complete information
about how it wants to be displayed.  The page itself contains
complete information on how it wants to be displayed.  Browsers
just aren't listening.  Complain to Netscape, not www-html.

A different, but related, issue, is how to display a presentation
on a machine less capable than that on which it was designed.  To
me, that just seems like the same old resource variant problem,
and so should be solved with existing methods, i.e. HTTP content
negotiation and nested <object> declarations.  Doing this will
require some additional information to standardize the format of,
so we could reasonably talk about that.

How about something like:

<object type="image/x-png; bitdepth=24" href="true-color.png">
   <object type="image/x-png; bitdepth=8" href="256-color.png">
     <img href="16-color.gif">
   </object>
</object>

and similarly for Accept: headers in HTTP?  Some indication of
monochrome vs. color might be useful as well.

I don't think we can sweep this under a rug by telling everyone
to get true color or go away; 8-bit displays will always be less
expensive than true-color because less RAM.  And monochrome will
always be less than color, so those displays will always be around.
There's no reason we should allow a designer to optimize for all
cases if he wants to.  It's even less acceptable to tell designers
to "design to the medium" by ignoring those who want full color
and sound if they've got it.  Advertising dollars speak louder
than pontifications on purity any day.